ARGUS - Aerial Rescue and Geospatial Utility System

Note

NEW VERSION IN DEVELOPMENT

Development of ARGUS 1.0 (main branch) has been discontinued – The current version can be found in the branch major-rewrite. ARGUS 2.0 is in active Development. Most features of ARGUS 1.0 are already available. As soon as all features are available in ARGUS 2.0, it will be released in the main branch. You are welcome to try ARGUS 2.0 with its improved UI and more modern architecture.

ARGUS is a documentation and analysis app designed for structured work with drone images in rescue operations. Its main functions include creating orthophotos from mapping flights with UAVs, presenting flight data and individual images in a structured manner, and evaluating infrared images. Additionally, ARGUS offers extensive object recognition functionalities with specially developed classifiers based on our own datasets (paper link, dataset link) from real-world missions. The established drone software OpenDroneMaps (ODM link ) is also integrated into ARGUS. Since ARGUS is a WebApp, it can be accessed from every device within the same network as the server. It is recommended to use Chrome (or any Chrome-based browser). Currently, Argus has only been tested with data from DJI drones (Matrice M30T, Mavic Enterprise, Mavic 2, Mavic 3). Using data from other drones can lead to problems.

As a new feature argus can now process 360° videos and reconstruct camera paths and partial point clouds, as well as giving you a 360 panoramic tour similar to google street view. The user can also upload two separate videos (e.g., in Insta360 format) instead of a pre-stitched 360° video. These will then be stitched together into a 360° video. However, it is recommended to perform this step in advance using the manufacturer's software for the 360° camera, as this typically provides faster and more reliable results. To speed up 360° video processing, increase the frame skip value (a value of 3 is a good starting point, adjustable in the 360° report settings), but note it may cause reconstruction errors. A 360 Report can currently only be created in the unsorted reports section. Keep in mind, that the 360 Report is not as polished as the other reports. A sample Video for testing is linked down below in the Example section.

This WebApp was developed at the Westphalian University of Applied Sciences (Westfälische Hochschule) as part of the E-DRZ research project, funded by the German Federal Ministry of Education and Research. For more details about our latest research findings, you can read our paper published at IEEE International Symposium on Safety, Security, and Rescue Robotics (SSRR2023), Fukushima, Japan, 13-15. Nov. 2023.

Please note that ARGUS is intended for use in a scientific context and does not offer the reliability and stability of fully developed commercial software.

Dependencies

To run the server, the following is needed:

• Linux
• Docker & Docker Compose (installation guide)
• (Optional) Nvidia Docker (installation guide)

Without Nvidia Docker, the object detection pipeline will use the CPU to run our neural networks, and WebODM will also run without the GPU.

Installing

1. Clone the repository: git clone --recursive https://github.com/RoblabWh/argus.git
2. Run ./install.sh from the repository directory. # take a coffee, building all images need up to 20 minutes
3. In case something goes wrong, try git submodule update --init --recursive

Running with Installation

1. Run argus up: this will create and start all the containers

Running without Installation

1. Clone the repository. Make sure to use --recursive or --recurse-submodules to also get the detection submodule and stella vslam dense submodule.
2. Navigate to the argus directory.
3. Run ./argus.sh up --build

Utilizing the systemd Service

1. Install ARGUS.
2. Enable the service with systemctl enable argus.

Uninstalling

Without Removing Images and User Data

Run /opt/argus/uninstall.sh.

With Removing Images and User Data

Run /opt/argus/uninstall.sh all.

Accessing the WebApp

The WebApp typically runs on port 5000. You can use either your localhost or the IP address of the machine running the WebApp. It should be visible to all devices on the same network.

Example

We have an already processed example project of a flooding in Germany in 2021 available (ArgusExample01.zip). Use the import feature under "new report" on the projects overview page to load the project.

A 360° video for testing the new feature can be found here (processing took a little bit more than 5 minutes with frame skip set to 3, on a Ryzen 7 5800X, with 32GB of RAM and an NVIDIA RTX 3080).

Short Demo video on YouTube: ARGUS - Aerial Rescue and Geospatial Utility System

Papers

1. Redefining Recon: Bridging Gaps with UAVs, 360° Cameras, and Neural Radiance Fields; Hartmut Surmann, Niklas Digakis, Jan-Nicklas Kremer, Julien Meine, Max Schulte, Niklas Voigt, IEEE International Symposium on Safety, Security, and Rescue Robotics (SSRR2023), Fukushima, Japan, 13-15. Nov. 2023 (cite this).
2. UAVs and Neural Networks for search and rescue missions, Hartmut Surmann, Artur Leinweber, Gerhard Senkowski, Julien Meine, Dominik Slomma, 56th International Symposium on Robotics (ISR Europe) September 26-27, 2023.
3. Lessons from Robot-Assisted Disaster Response Deployments by the German Rescue Robotics Center Task Force, Hartmut Surmann, Ivana Kruijff-Korbayova, Kevin Daun, Marius Schnaubelt, Oskar von Stryk, Manuel Patchou, Stefan Boecker, Christian Wietfeld, Jan Quenzel, Daniel Schleich, Sven Behnke, Robert Grafe, Nils Heidemann, Dominik Slomma, Journal of Field robotics, Wiley, 2023.
4. Deployment of Aerial Robots during the Flood Disaster in Erftstadt / Blessem in July 2021, Hartmut Surmann, Dominik Slomma, Robert Grafe, Stefan Grobelny, 2022 8th International Conference on Automation, Robotics and Applications (ICARA), Prague, Czech Republic, 2022, pp. 97-102.
5. Deployment of Aerial Robots after a major fire of an industrial hall with hazardous substances, a report, Hartmut Surmann, Dominik Slomma, Stefan Grobelny, Robert Grafe, 2021 IEEE International Symposium on Safety, Security, and Rescue Robotics (SSRR2021), New York City, NY, USA, 2021, pp. 40-47.

Known Issues

• Firefox may encounter problems when uploading larger files, such as high-resolution panoramic photos.
• Since ARGUS is still in development, starting multiple tasks simultaneously can lead to unexpected behavior or, in rare cases, system crashes.
• Currently, ARGUS primarily supports and is tested with DJI drones (DJI M30T, as well as multiple Mavic and Mavic Enterprise models).
• In order to generate fast orthophotos, the UAVs gimbal should orient the camera orthogonal towards ground (-90°).
• If WebODM has many 'WARNING Bad Request: /api/token-auth/' during startup and the redirection from ARGUS does not work, try pulling a new WebODM image with './argus.sh pull --ignore-buildable' and rebuild the server with './argus.sh up --build'.
• Some older Linus operating systems may have problems with calling docker compose without a hyphen in between (docker-compose instead of docker compose).